Automated computer analysis of the electrocardiogram

Interpretation of the electrocardiogram (ECG) has become one of the larger tasks now being placed on the cardiologist. Besides the magnitude of the task, variability among cardiologists and aid for the rural population are two other reasons for looking toward automation in an effort to reduce the burden placed on the cardiologist. This thesis presents the work done in developing computer programs to automate the analysis of the ECGo. Three phases of ECG analysis must be considered. The first is analysis of the contour of the ECG complex in order to detect abnormalities in the electrical conduction system of the heart muscle. The second phase is analysis of the heart rhythms, i.e., the relationship between ventricular depolarizations and atrial depolarizations. The final phase is an analysis of the changes which take place in the ECG over a period of time. Programs to tackle each of these problems have been developed and are described in the thesis. Two problems exist in developing computer ECG analysis programs. The first is electrical "noise" which is superimposed on the signa1. The other problem is development of precise logic for proper interpretation of the ECG. Each of these problems are discussed with solutions given to a portion of the difficulties. The contour and rhythm programs were tested against 202 ECGls which had been interpreted by one or more cardiologists. Detailed results of the study are included in the thesis. These results showed 8% false positives and 2% false negatives. The sequential analysis program was evaluated using 50 pairs of ECG's. It was found that the program detected 22 ECG's which showed minor changes and 6 ECG's which showed major changes. The future of such programs are considered with the type of research which is necessary to overcome the present difficulties in the program.